Polymorphisms in the ICAM Gene Locus are not Associated with Breast Cancer Risk

In a recent case-control study which examined >25,000 single nucleotide polymorphisms (SNP), the intercellular adhesion molecule (ICAM) locus was proposed as a susceptibility locus for breast and prostate cancer (1). The genes ICAM1, ICAM4, and ICAM5 are all located within a 20 kb region of high linkage disequilibrium of chromosome 19p13.2. We examined the association of three previously implicated polymorphisms (rs5030382, K469E in ICAM1; rs281439, 542 bp upstream of ICAM5; rs1056538, and V301I in ICAM5) with breast cancer risk in a nested case control study within the Nurses' Health Study.

Genotyping assays for the ICAM locus polymorphisms (rs5030382, K469E in ICAM1; rs281439, 542 bp upstream of ICAM5; rs1056538, V301I in ICAM5) were done by the 5′-nuclease assay (TaqMan) on the ABI Prism 7900HT Sequence Detection System (Applied Biosystems, Foster City, CA). Information on TaqMan genotyping assays is available on request from the authors. Our study included a total of 1,264 incident breast cancer cases diagnosed after blood draw up to June 1, 2000, and 1,747 matched controls, all drawn from 32,826 women who gave a blood sample from 1989 to 1990. Controls were randomly selected participants who were free of diagnosed cancer (except non–melanoma skin cancer), and matched to cases based on age, menopausal status, recent postmenopausal hormone use, and time, day, and month of blood collection. A detailed description of this study population has been previously reported (2). Approximately 95% of the samples were successfully genotyped; samples that failed genotyping were removed from the analyses. Internal blinded quality control samples showed 100% concordance.

We used SAS v8.2 (SAS Institute, Cary, NC) for all statistical analyses. Odds ratios (OR) and 95% confidence intervals were calculated using unconditional logistic regression, controlling for matching factors, age at menopause, age at menarche, age at first birth and parity, history of benign breast disease, and family history of breast cancer using PROC LOGISTIC. We tested for departures from Hardy-Weinberg equilibrium using PROC ALLELE. Interactions were tested by likelihood ratio tests comparing the model with main effects for each variable of interest to the model additionally including the two variables cross-tabulated. All P values reported are two-sided. Power calculations were carried out using Quanto (3).

No departure from Hardy-Weinberg equilibrium was observed at any of the SNPs examined in the control population. We observed no statistically significant differences in allele or genotype frequencies between cases and controls; conditional, unconditional and unconditional multivariate logistic regressions gave similar results, with ORs not significantly different from 1.00 (Table 1). No difference in genotype frequencies were observed between cases with estrogen or progesterone receptor–positive and estrogen or progesterone receptor–negative tumors or other tumor characteristic groups, such as tumor size and grade (data not shown). No statistically significant interactions were observed between family history of breast cancer or age at diagnosis and the polymorphisms assayed (data not shown).

ICAM1 has been proposed as a likely candidate for genetic susceptibility to breast cancer. Soluble levels of ICAM-1 in the sera of patients with stage IV breast cancer were higher than that of healthy controls (4) and patients with lower grade tumors (5). ICAM-1 is hypothesized to be involved in the adhesion of tumor cells to the vascular epithelium, and therefore, promote the development of metastases.

In a large-scale association study of >25,000 SNPs in ∼16,000 genes, Kammerer et al. (1) identified the ICAM locus at chromosome 19p13.2 as a region of susceptibility to breast cancer. They used one small study of Germans (254 cases and 268 controls) as a “discovery” set, and two smaller studies for replication (188 cases and 150 controls from Germany, 180 cases and 180 controls from Australia). Only the ICAM5 V301I polymorphism showed a statistically significant association (P = 0.001 for difference in allele frequency between cases and controls) with breast cancer risk in the discovery set, which was marginally supported by the replication sets (P = 0.07 and 0.03, respectively).

If the three studies reported by Kammerer et al. are combined, statistically significant associations between the ICAM5 V301I and ICAM1 K469E polymorphisms exist. Using the data they report, an OR of 0.55 for the ICAM5 V301I AA versus GG, and an OR of 0.63 for the ICAM1 K469E GG versus AA could be calculated. Our study has >93% power to detect log-additive OR < 0.81 with SNPs similar in allele frequency to those reported here at the type 1 error rate of 1%. The allele frequencies between our study and those of Kammerer et al. were similar for all three SNPs among the control populations.

Our study does not confirm the associations observed by Kammerer et al. Cases and controls in our study and those reported by Kammerer et al. are predominantly of European origin. Slight differences in age at diagnosis (median of 63 years in our study, opposed to 56, 50, and 65 in Kammerer et al.) exist between these studies. Control samples in our study were not excluded based on family history of breast cancer, although they were excluded from the German discovery set and the Australian replication set of Kammerer et al. However, limiting our control group to those with no family history of breast cancer (1,490 controls) did not change the results (data not shown). Lastly, our study is based on prospective samples, with blood collected before cancer diagnosis. These differences, however, are not likely explanations for the differences in risk estimates between the two studies. The differences in risk estimates between these studies are more than likely due to chance variation in genotype distributions. The large sample size of our study limits the possibility that this is a false-negative result.

The cost of genotyping is steadily declining, whereas the speed of genotyping is steadily increasing. This is making it more feasible to examine many loci in large sample sets. However, in order to avoid false-positive findings, both discovery and replication studies of sufficient size need to be carried out.